fluke.distr

This module contains classes and functions for parallel federated learning using multiple GPUs.

Submodules

client
server
utils	Utility module for fluke.distr.

Classes included in fluke.distr

ParallelAlgorithm

ParallelAlgorithm is a class for parallel federated learning algorithms that use multiple GPUs.

class fluke.distr.ParallelAlgorithm

class fluke.distr.ParallelAlgorithm(n_clients: int, data_splitter: DataSplitter, hyper_params: DDict | dict[str, Any], clients: list[ParallelClient] = None, server: ParallelServer = None, **kwargs)

ParallelAlgorithm is a class for parallel federated learning algorithms that use multiple GPUs. It inherits from CentralizedFL and initializes clients and server for parallel execution. Clients training runs on different GPUs, and the server aggregates their updates.